net/mptcp/pm_netlink.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /* Multipath TCP
   3  *
   4  * Copyright (c) 2020, Red Hat, Inc.
   5  */
   6
   7 #define pr_fmt(fmt) "MPTCP: " fmt
   8
   9 #include <linux/inet.h>
  10 #include <linux/kernel.h>
  11 #include <net/tcp.h>
  12 #include <net/netns/generic.h>
  13 #include <net/mptcp.h>
  14 #include <net/genetlink.h>
  15 #include <uapi/linux/mptcp.h>
  16
  17 #include "protocol.h"
  18 #include "mib.h"
  19
  20 /* forward declaration */
  21 static struct genl_family mptcp_genl_family;
  22
  23 static int pm_nl_pernet_id;
  24
  25 struct mptcp_pm_addr_entry {
  26         struct list_head        list;
  27         struct mptcp_addr_info  addr;
  28         u8                      flags;
  29         int                     ifindex;
  30         struct rcu_head         rcu;
  31         struct socket           *lsk;
  32 };
  33
  34 struct mptcp_pm_add_entry {
  35         struct list_head        list;
  36         struct mptcp_addr_info  addr;
  37         struct timer_list       add_timer;
  38         struct mptcp_sock       *sock;
  39         u8                      retrans_times;
  40 };
  41
  42 #define MAX_ADDR_ID             255
  43 #define BITMAP_SZ DIV_ROUND_UP(MAX_ADDR_ID + 1, BITS_PER_LONG)
  44
  45 struct pm_nl_pernet {
  46         /* protects pernet updates */
  47         spinlock_t              lock;
  48         struct list_head        local_addr_list;
  49         unsigned int            addrs;
  50         unsigned int            add_addr_signal_max;
  51         unsigned int            add_addr_accept_max;
  52         unsigned int            local_addr_max;
  53         unsigned int            subflows_max;
  54         unsigned int            next_id;
  55         unsigned long           id_bitmap[BITMAP_SZ];
  56 };
  57
  58 #define MPTCP_PM_ADDR_MAX       8
  59 #define ADD_ADDR_RETRANS_MAX    3
  60
  61 static bool addresses_equal(const struct mptcp_addr_info *a,
  62                             struct mptcp_addr_info *b, bool use_port)
  63 {
  64         bool addr_equals = false;
  65
  66         if (a->family == b->family) {
  67                 if (a->family == AF_INET)
  68                         addr_equals = a->addr.s_addr == b->addr.s_addr;
  69 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
  70                 else
  71                         addr_equals = !ipv6_addr_cmp(&a->addr6, &b->addr6);
  72         } else if (a->family == AF_INET) {
  73                 if (ipv6_addr_v4mapped(&b->addr6))
  74                         addr_equals = a->addr.s_addr == b->addr6.s6_addr32[3];
  75         } else if (b->family == AF_INET) {
  76                 if (ipv6_addr_v4mapped(&a->addr6))
  77                         addr_equals = a->addr6.s6_addr32[3] == b->addr.s_addr;
  78 #endif
  79         }
  80
  81         if (!addr_equals)
  82                 return false;
  83         if (!use_port)
  84                 return true;
  85
  86         return a->port == b->port;
  87 }
  88
  89 static bool address_zero(const struct mptcp_addr_info *addr)
  90 {
  91         struct mptcp_addr_info zero;
  92
  93         memset(&zero, 0, sizeof(zero));
  94         zero.family = addr->family;
  95
  96         return addresses_equal(addr, &zero, true);
  97 }
  98
  99 static void local_address(const struct sock_common *skc,
 100                           struct mptcp_addr_info *addr)
 101 {
 102         addr->family = skc->skc_family;
 103         addr->port = htons(skc->skc_num);
 104         if (addr->family == AF_INET)
 105                 addr->addr.s_addr = skc->skc_rcv_saddr;
 106 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 107         else if (addr->family == AF_INET6)
 108                 addr->addr6 = skc->skc_v6_rcv_saddr;
 109 #endif
 110 }
 111
 112 static void remote_address(const struct sock_common *skc,
 113                            struct mptcp_addr_info *addr)
 114 {
 115         addr->family = skc->skc_family;
 116         addr->port = skc->skc_dport;
 117         if (addr->family == AF_INET)
 118                 addr->addr.s_addr = skc->skc_daddr;
 119 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 120         else if (addr->family == AF_INET6)
 121                 addr->addr6 = skc->skc_v6_daddr;
 122 #endif
 123 }
 124
 125 static bool lookup_subflow_by_saddr(const struct list_head *list,
 126                                     struct mptcp_addr_info *saddr)
 127 {
 128         struct mptcp_subflow_context *subflow;
 129         struct mptcp_addr_info cur;
 130         struct sock_common *skc;
 131
 132         list_for_each_entry(subflow, list, node) {
 133                 skc = (struct sock_common *)mptcp_subflow_tcp_sock(subflow);
 134
 135                 local_address(skc, &cur);
 136                 if (addresses_equal(&cur, saddr, saddr->port))
 137                         return true;
 138         }
 139
 140         return false;
 141 }
 142
 143 static bool lookup_subflow_by_daddr(const struct list_head *list,
 144                                     struct mptcp_addr_info *daddr)
 145 {
 146         struct mptcp_subflow_context *subflow;
 147         struct mptcp_addr_info cur;
 148         struct sock_common *skc;
 149
 150         list_for_each_entry(subflow, list, node) {
 151                 skc = (struct sock_common *)mptcp_subflow_tcp_sock(subflow);
 152
 153                 remote_address(skc, &cur);
 154                 if (addresses_equal(&cur, daddr, daddr->port))
 155                         return true;
 156         }
 157
 158         return false;
 159 }
 160
 161 static struct mptcp_pm_addr_entry *
 162 select_local_address(const struct pm_nl_pernet *pernet,
 163                      struct mptcp_sock *msk)
 164 {
 165         struct mptcp_pm_addr_entry *entry, *ret = NULL;
 166         struct sock *sk = (struct sock *)msk;
 167
 168         msk_owned_by_me(msk);
 169
 170         rcu_read_lock();
 171         __mptcp_flush_join_list(msk);
 172         list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
 173                 if (!(entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW))
 174                         continue;
 175
 176                 if (entry->addr.family != sk->sk_family) {
 177 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 178                         if ((entry->addr.family == AF_INET &&
 179                              !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) ||
 180                             (sk->sk_family == AF_INET &&
 181                              !ipv6_addr_v4mapped(&entry->addr.addr6)))
 182 #endif
 183                                 continue;
 184                 }
 185
 186                 /* avoid any address already in use by subflows and
 187                  * pending join
 188                  */
 189                 if (!lookup_subflow_by_saddr(&msk->conn_list, &entry->addr)) {
 190                         ret = entry;
 191                         break;
 192                 }
 193         }
 194         rcu_read_unlock();
 195         return ret;
 196 }
 197
 198 static struct mptcp_pm_addr_entry *
 199 select_signal_address(struct pm_nl_pernet *pernet, unsigned int pos)
 200 {
 201         struct mptcp_pm_addr_entry *entry, *ret = NULL;
 202         int i = 0;
 203
 204         rcu_read_lock();
 205         /* do not keep any additional per socket state, just signal
 206          * the address list in order.
 207          * Note: removal from the local address list during the msk life-cycle
 208          * can lead to additional addresses not being announced.
 209          */
 210         list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
 211                 if (!(entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL))
 212                         continue;
 213                 if (i++ == pos) {
 214                         ret = entry;
 215                         break;
 216                 }
 217         }
 218         rcu_read_unlock();
 219         return ret;
 220 }
 221
 222 unsigned int mptcp_pm_get_add_addr_signal_max(struct mptcp_sock *msk)
 223 {
 224         struct pm_nl_pernet *pernet;
 225
 226         pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
 227         return READ_ONCE(pernet->add_addr_signal_max);
 228 }
 229 EXPORT_SYMBOL_GPL(mptcp_pm_get_add_addr_signal_max);
 230
 231 unsigned int mptcp_pm_get_add_addr_accept_max(struct mptcp_sock *msk)
 232 {
 233         struct pm_nl_pernet *pernet;
 234
 235         pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
 236         return READ_ONCE(pernet->add_addr_accept_max);
 237 }
 238 EXPORT_SYMBOL_GPL(mptcp_pm_get_add_addr_accept_max);
 239
 240 unsigned int mptcp_pm_get_subflows_max(struct mptcp_sock *msk)
 241 {
 242         struct pm_nl_pernet *pernet;
 243
 244         pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
 245         return READ_ONCE(pernet->subflows_max);
 246 }
 247 EXPORT_SYMBOL_GPL(mptcp_pm_get_subflows_max);
 248
 249 unsigned int mptcp_pm_get_local_addr_max(struct mptcp_sock *msk)
 250 {
 251         struct pm_nl_pernet *pernet;
 252
 253         pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
 254         return READ_ONCE(pernet->local_addr_max);
 255 }
 256 EXPORT_SYMBOL_GPL(mptcp_pm_get_local_addr_max);
 257
 258 static void check_work_pending(struct mptcp_sock *msk)
 259 {
 260         if (msk->pm.add_addr_signaled == mptcp_pm_get_add_addr_signal_max(msk) &&
 261             (msk->pm.local_addr_used == mptcp_pm_get_local_addr_max(msk) ||
 262              msk->pm.subflows == mptcp_pm_get_subflows_max(msk)))
 263                 WRITE_ONCE(msk->pm.work_pending, false);
 264 }
 265
 266 struct mptcp_pm_add_entry *
 267 mptcp_lookup_anno_list_by_saddr(struct mptcp_sock *msk,
 268                                 struct mptcp_addr_info *addr)
 269 {
 270         struct mptcp_pm_add_entry *entry;
 271
 272         lockdep_assert_held(&msk->pm.lock);
 273
 274         list_for_each_entry(entry, &msk->pm.anno_list, list) {
 275                 if (addresses_equal(&entry->addr, addr, true))
 276                         return entry;
 277         }
 278
 279         return NULL;
 280 }
 281
 282 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk)
 283 {
 284         struct mptcp_pm_add_entry *entry;
 285         struct mptcp_addr_info saddr;
 286         bool ret = false;
 287
 288         local_address((struct sock_common *)sk, &saddr);
 289
 290         spin_lock_bh(&msk->pm.lock);
 291         list_for_each_entry(entry, &msk->pm.anno_list, list) {
 292                 if (addresses_equal(&entry->addr, &saddr, true)) {
 293                         ret = true;
 294                         goto out;
 295                 }
 296         }
 297
 298 out:
 299         spin_unlock_bh(&msk->pm.lock);
 300         return ret;
 301 }
 302
 303 static void mptcp_pm_add_timer(struct timer_list *timer)
 304 {
 305         struct mptcp_pm_add_entry *entry = from_timer(entry, timer, add_timer);
 306         struct mptcp_sock *msk = entry->sock;
 307         struct sock *sk = (struct sock *)msk;
 308
 309         pr_debug("msk=%p", msk);
 310
 311         if (!msk)
 312                 return;
 313
 314         if (inet_sk_state_load(sk) == TCP_CLOSE)
 315                 return;
 316
 317         if (!entry->addr.id)
 318                 return;
 319
 320         if (mptcp_pm_should_add_signal(msk)) {
 321                 sk_reset_timer(sk, timer, jiffies + TCP_RTO_MAX / 8);
 322                 goto out;
 323         }
 324
 325         spin_lock_bh(&msk->pm.lock);
 326
 327         if (!mptcp_pm_should_add_signal(msk)) {
 328                 pr_debug("retransmit ADD_ADDR id=%d", entry->addr.id);
 329                 mptcp_pm_announce_addr(msk, &entry->addr, false);
 330                 mptcp_pm_add_addr_send_ack(msk);
 331                 entry->retrans_times++;
 332         }
 333
 334         if (entry->retrans_times < ADD_ADDR_RETRANS_MAX)
 335                 sk_reset_timer(sk, timer,
 336                                jiffies + mptcp_get_add_addr_timeout(sock_net(sk)));
 337
 338         spin_unlock_bh(&msk->pm.lock);
 339
 340         if (entry->retrans_times == ADD_ADDR_RETRANS_MAX)
 341                 mptcp_pm_subflow_established(msk);
 342
 343 out:
 344         __sock_put(sk);
 345 }
 346
 347 struct mptcp_pm_add_entry *
 348 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
 349                        struct mptcp_addr_info *addr, bool check_id)
 350 {
 351         struct mptcp_pm_add_entry *entry;
 352         struct sock *sk = (struct sock *)msk;
 353
 354         spin_lock_bh(&msk->pm.lock);
 355         entry = mptcp_lookup_anno_list_by_saddr(msk, addr);
 356         if (entry && (!check_id || entry->addr.id == addr->id))
 357                 entry->retrans_times = ADD_ADDR_RETRANS_MAX;
 358         spin_unlock_bh(&msk->pm.lock);
 359
 360         if (entry && (!check_id || entry->addr.id == addr->id))
 361                 sk_stop_timer_sync(sk, &entry->add_timer);
 362
 363         return entry;
 364 }
 365
 366 static bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
 367                                      struct mptcp_pm_addr_entry *entry)
 368 {
 369         struct mptcp_pm_add_entry *add_entry = NULL;
 370         struct sock *sk = (struct sock *)msk;
 371         struct net *net = sock_net(sk);
 372
 373         lockdep_assert_held(&msk->pm.lock);
 374
 375         if (mptcp_lookup_anno_list_by_saddr(msk, &entry->addr))
 376                 return false;
 377
 378         add_entry = kmalloc(sizeof(*add_entry), GFP_ATOMIC);
 379         if (!add_entry)
 380                 return false;
 381
 382         list_add(&add_entry->list, &msk->pm.anno_list);
 383
 384         add_entry->addr = entry->addr;
 385         add_entry->sock = msk;
 386         add_entry->retrans_times = 0;
 387
 388         timer_setup(&add_entry->add_timer, mptcp_pm_add_timer, 0);
 389         sk_reset_timer(sk, &add_entry->add_timer,
 390                        jiffies + mptcp_get_add_addr_timeout(net));
 391
 392         return true;
 393 }
 394
 395 void mptcp_pm_free_anno_list(struct mptcp_sock *msk)
 396 {
 397         struct mptcp_pm_add_entry *entry, *tmp;
 398         struct sock *sk = (struct sock *)msk;
 399         LIST_HEAD(free_list);
 400
 401         pr_debug("msk=%p", msk);
 402
 403         spin_lock_bh(&msk->pm.lock);
 404         list_splice_init(&msk->pm.anno_list, &free_list);
 405         spin_unlock_bh(&msk->pm.lock);
 406
 407         list_for_each_entry_safe(entry, tmp, &free_list, list) {
 408                 sk_stop_timer_sync(sk, &entry->add_timer);
 409                 kfree(entry);
 410         }
 411 }
 412
 413 static void mptcp_pm_create_subflow_or_signal_addr(struct mptcp_sock *msk)
 414 {
 415         struct sock *sk = (struct sock *)msk;
 416         struct mptcp_pm_addr_entry *local;
 417         unsigned int add_addr_signal_max;
 418         unsigned int local_addr_max;
 419         struct pm_nl_pernet *pernet;
 420         unsigned int subflows_max;
 421
 422         pernet = net_generic(sock_net(sk), pm_nl_pernet_id);
 423
 424         add_addr_signal_max = mptcp_pm_get_add_addr_signal_max(msk);
 425         local_addr_max = mptcp_pm_get_local_addr_max(msk);
 426         subflows_max = mptcp_pm_get_subflows_max(msk);
 427
 428         pr_debug("local %d:%d signal %d:%d subflows %d:%d\n",
 429                  msk->pm.local_addr_used, local_addr_max,
 430                  msk->pm.add_addr_signaled, add_addr_signal_max,
 431                  msk->pm.subflows, subflows_max);
 432
 433         /* check first for announce */
 434         if (msk->pm.add_addr_signaled < add_addr_signal_max) {
 435                 local = select_signal_address(pernet,
 436                                               msk->pm.add_addr_signaled);
 437
 438                 if (local) {
 439                         if (mptcp_pm_alloc_anno_list(msk, local)) {
 440                                 msk->pm.add_addr_signaled++;
 441                                 mptcp_pm_announce_addr(msk, &local->addr, false);
 442                                 mptcp_pm_nl_addr_send_ack(msk);
 443                         }
 444                 } else {
 445                         /* pick failed, avoid fourther attempts later */
 446                         msk->pm.local_addr_used = add_addr_signal_max;
 447                 }
 448
 449                 check_work_pending(msk);
 450         }
 451
 452         /* check if should create a new subflow */
 453         if (msk->pm.local_addr_used < local_addr_max &&
 454             msk->pm.subflows < subflows_max &&
 455             !READ_ONCE(msk->pm.remote_deny_join_id0)) {
 456                 local = select_local_address(pernet, msk);
 457                 if (local) {
 458                         struct mptcp_addr_info remote = { 0 };
 459
 460                         msk->pm.local_addr_used++;
 461                         msk->pm.subflows++;
 462                         check_work_pending(msk);
 463                         remote_address((struct sock_common *)sk, &remote);
 464                         spin_unlock_bh(&msk->pm.lock);
 465                         __mptcp_subflow_connect(sk, &local->addr, &remote,
 466                                                 local->flags, local->ifindex);
 467                         spin_lock_bh(&msk->pm.lock);
 468                         return;
 469                 }
 470
 471                 /* lookup failed, avoid fourther attempts later */
 472                 msk->pm.local_addr_used = local_addr_max;
 473                 check_work_pending(msk);
 474         }
 475 }
 476
 477 static void mptcp_pm_nl_fully_established(struct mptcp_sock *msk)
 478 {
 479         mptcp_pm_create_subflow_or_signal_addr(msk);
 480 }
 481
 482 static void mptcp_pm_nl_subflow_established(struct mptcp_sock *msk)
 483 {
 484         mptcp_pm_create_subflow_or_signal_addr(msk);
 485 }
 486
 487 static void mptcp_pm_nl_add_addr_received(struct mptcp_sock *msk)
 488 {
 489         struct sock *sk = (struct sock *)msk;
 490         unsigned int add_addr_accept_max;
 491         struct mptcp_addr_info remote;
 492         struct mptcp_addr_info local;
 493         unsigned int subflows_max;
 494
 495         add_addr_accept_max = mptcp_pm_get_add_addr_accept_max(msk);
 496         subflows_max = mptcp_pm_get_subflows_max(msk);
 497
 498         pr_debug("accepted %d:%d remote family %d",
 499                  msk->pm.add_addr_accepted, add_addr_accept_max,
 500                  msk->pm.remote.family);
 501
 502         if (lookup_subflow_by_daddr(&msk->conn_list, &msk->pm.remote))
 503                 goto add_addr_echo;
 504
 505         msk->pm.add_addr_accepted++;
 506         msk->pm.subflows++;
 507         if (msk->pm.add_addr_accepted >= add_addr_accept_max ||
 508             msk->pm.subflows >= subflows_max)
 509                 WRITE_ONCE(msk->pm.accept_addr, false);
 510
 511         /* connect to the specified remote address, using whatever
 512          * local address the routing configuration will pick.
 513          */
 514         remote = msk->pm.remote;
 515         if (!remote.port)
 516                 remote.port = sk->sk_dport;
 517         memset(&local, 0, sizeof(local));
 518         local.family = remote.family;
 519
 520         spin_unlock_bh(&msk->pm.lock);
 521         __mptcp_subflow_connect(sk, &local, &remote, 0, 0);
 522         spin_lock_bh(&msk->pm.lock);
 523
 524 add_addr_echo:
 525         mptcp_pm_announce_addr(msk, &msk->pm.remote, true);
 526         mptcp_pm_nl_addr_send_ack(msk);
 527 }
 528
 529 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk)
 530 {
 531         struct mptcp_subflow_context *subflow;
 532
 533         msk_owned_by_me(msk);
 534         lockdep_assert_held(&msk->pm.lock);
 535
 536         if (!mptcp_pm_should_add_signal(msk) &&
 537             !mptcp_pm_should_rm_signal(msk))
 538                 return;
 539
 540         __mptcp_flush_join_list(msk);
 541         subflow = list_first_entry_or_null(&msk->conn_list, typeof(*subflow), node);
 542         if (subflow) {
 543                 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
 544                 bool slow;
 545
 546                 spin_unlock_bh(&msk->pm.lock);
 547                 pr_debug("send ack for %s%s%s",
 548                          mptcp_pm_should_add_signal(msk) ? "add_addr" : "rm_addr",
 549                          mptcp_pm_should_add_signal_ipv6(msk) ? " [ipv6]" : "",
 550                          mptcp_pm_should_add_signal_port(msk) ? " [port]" : "");
 551
 552                 slow = lock_sock_fast(ssk);
 553                 tcp_send_ack(ssk);
 554                 unlock_sock_fast(ssk, slow);
 555                 spin_lock_bh(&msk->pm.lock);
 556         }
 557 }
 558
 559 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
 560                                  struct mptcp_addr_info *addr,
 561                                  u8 bkup)
 562 {
 563         struct mptcp_subflow_context *subflow;
 564
 565         pr_debug("bkup=%d", bkup);
 566
 567         mptcp_for_each_subflow(msk, subflow) {
 568                 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
 569                 struct sock *sk = (struct sock *)msk;
 570                 struct mptcp_addr_info local;
 571                 bool slow;
 572
 573                 local_address((struct sock_common *)ssk, &local);
 574                 if (!addresses_equal(&local, addr, addr->port))
 575                         continue;
 576
 577                 subflow->backup = bkup;
 578                 subflow->send_mp_prio = 1;
 579                 subflow->request_bkup = bkup;
 580                 __MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_MPPRIOTX);
 581
 582                 spin_unlock_bh(&msk->pm.lock);
 583                 pr_debug("send ack for mp_prio");
 584                 slow = lock_sock_fast(ssk);
 585                 tcp_send_ack(ssk);
 586                 unlock_sock_fast(ssk, slow);
 587                 spin_lock_bh(&msk->pm.lock);
 588
 589                 return 0;
 590         }
 591
 592         return -EINVAL;
 593 }
 594
 595 static void mptcp_pm_nl_rm_addr_or_subflow(struct mptcp_sock *msk,
 596                                            const struct mptcp_rm_list *rm_list,
 597                                            enum linux_mptcp_mib_field rm_type)
 598 {
 599         struct mptcp_subflow_context *subflow, *tmp;
 600         struct sock *sk = (struct sock *)msk;
 601         u8 i;
 602
 603         pr_debug("%s rm_list_nr %d",
 604                  rm_type == MPTCP_MIB_RMADDR ? "address" : "subflow", rm_list->nr);
 605
 606         msk_owned_by_me(msk);
 607
 608         if (!rm_list->nr)
 609                 return;
 610
 611         if (list_empty(&msk->conn_list))
 612                 return;
 613
 614         for (i = 0; i < rm_list->nr; i++) {
 615                 list_for_each_entry_safe(subflow, tmp, &msk->conn_list, node) {
 616                         struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
 617                         int how = RCV_SHUTDOWN | SEND_SHUTDOWN;
 618                         u8 id = subflow->local_id;
 619
 620                         if (rm_type == MPTCP_MIB_RMADDR)
 621                                 id = subflow->remote_id;
 622
 623                         if (rm_list->ids[i] != id)
 624                                 continue;
 625
 626                         pr_debug(" -> %s rm_list_ids[%d]=%u local_id=%u remote_id=%u",
 627                                  rm_type == MPTCP_MIB_RMADDR ? "address" : "subflow",
 628                                  i, rm_list->ids[i], subflow->local_id, subflow->remote_id);
 629                         spin_unlock_bh(&msk->pm.lock);
 630                         mptcp_subflow_shutdown(sk, ssk, how);
 631                         mptcp_close_ssk(sk, ssk, subflow);
 632                         spin_lock_bh(&msk->pm.lock);
 633
 634                         if (rm_type == MPTCP_MIB_RMADDR) {
 635                                 msk->pm.add_addr_accepted--;
 636                                 WRITE_ONCE(msk->pm.accept_addr, true);
 637                         } else if (rm_type == MPTCP_MIB_RMSUBFLOW) {
 638                                 msk->pm.local_addr_used--;
 639                         }
 640                         msk->pm.subflows--;
 641                         __MPTCP_INC_STATS(sock_net(sk), rm_type);
 642                 }
 643         }
 644 }
 645
 646 static void mptcp_pm_nl_rm_addr_received(struct mptcp_sock *msk)
 647 {
 648         mptcp_pm_nl_rm_addr_or_subflow(msk, &msk->pm.rm_list_rx, MPTCP_MIB_RMADDR);
 649 }
 650
 651 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
 652                                      const struct mptcp_rm_list *rm_list)
 653 {
 654         mptcp_pm_nl_rm_addr_or_subflow(msk, rm_list, MPTCP_MIB_RMSUBFLOW);
 655 }
 656
 657 void mptcp_pm_nl_work(struct mptcp_sock *msk)
 658 {
 659         struct mptcp_pm_data *pm = &msk->pm;
 660
 661         msk_owned_by_me(msk);
 662
 663         spin_lock_bh(&msk->pm.lock);
 664
 665         pr_debug("msk=%p status=%x", msk, pm->status);
 666         if (pm->status & BIT(MPTCP_PM_ADD_ADDR_RECEIVED)) {
 667                 pm->status &= ~BIT(MPTCP_PM_ADD_ADDR_RECEIVED);
 668                 mptcp_pm_nl_add_addr_received(msk);
 669         }
 670         if (pm->status & BIT(MPTCP_PM_ADD_ADDR_SEND_ACK)) {
 671                 pm->status &= ~BIT(MPTCP_PM_ADD_ADDR_SEND_ACK);
 672                 mptcp_pm_nl_addr_send_ack(msk);
 673         }
 674         if (pm->status & BIT(MPTCP_PM_RM_ADDR_RECEIVED)) {
 675                 pm->status &= ~BIT(MPTCP_PM_RM_ADDR_RECEIVED);
 676                 mptcp_pm_nl_rm_addr_received(msk);
 677         }
 678         if (pm->status & BIT(MPTCP_PM_ESTABLISHED)) {
 679                 pm->status &= ~BIT(MPTCP_PM_ESTABLISHED);
 680                 mptcp_pm_nl_fully_established(msk);
 681         }
 682         if (pm->status & BIT(MPTCP_PM_SUBFLOW_ESTABLISHED)) {
 683                 pm->status &= ~BIT(MPTCP_PM_SUBFLOW_ESTABLISHED);
 684                 mptcp_pm_nl_subflow_established(msk);
 685         }
 686
 687         spin_unlock_bh(&msk->pm.lock);
 688 }
 689
 690 static bool address_use_port(struct mptcp_pm_addr_entry *entry)
 691 {
 692         return (entry->flags &
 693                 (MPTCP_PM_ADDR_FLAG_SIGNAL | MPTCP_PM_ADDR_FLAG_SUBFLOW)) ==
 694                 MPTCP_PM_ADDR_FLAG_SIGNAL;
 695 }
 696
 697 static int mptcp_pm_nl_append_new_local_addr(struct pm_nl_pernet *pernet,
 698                                              struct mptcp_pm_addr_entry *entry)
 699 {
 700         struct mptcp_pm_addr_entry *cur;
 701         unsigned int addr_max;
 702         int ret = -EINVAL;
 703
 704         spin_lock_bh(&pernet->lock);
 705         /* to keep the code simple, don't do IDR-like allocation for address ID,
 706          * just bail when we exceed limits
 707          */
 708         if (pernet->next_id == MAX_ADDR_ID)
 709                 pernet->next_id = 1;
 710         if (pernet->addrs >= MPTCP_PM_ADDR_MAX)
 711                 goto out;
 712         if (test_bit(entry->addr.id, pernet->id_bitmap))
 713                 goto out;
 714
 715         /* do not insert duplicate address, differentiate on port only
 716          * singled addresses
 717          */
 718         list_for_each_entry(cur, &pernet->local_addr_list, list) {
 719                 if (addresses_equal(&cur->addr, &entry->addr,
 720                                     address_use_port(entry) &&
 721                                     address_use_port(cur)))
 722                         goto out;
 723         }
 724
 725         if (!entry->addr.id) {
 726 find_next:
 727                 entry->addr.id = find_next_zero_bit(pernet->id_bitmap,
 728                                                     MAX_ADDR_ID + 1,
 729                                                     pernet->next_id);
 730                 if ((!entry->addr.id || entry->addr.id > MAX_ADDR_ID) &&
 731                     pernet->next_id != 1) {
 732                         pernet->next_id = 1;
 733                         goto find_next;
 734                 }
 735         }
 736
 737         if (!entry->addr.id || entry->addr.id > MAX_ADDR_ID)
 738                 goto out;
 739
 740         __set_bit(entry->addr.id, pernet->id_bitmap);
 741         if (entry->addr.id > pernet->next_id)
 742                 pernet->next_id = entry->addr.id;
 743
 744         if (entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL) {
 745                 addr_max = pernet->add_addr_signal_max;
 746                 WRITE_ONCE(pernet->add_addr_signal_max, addr_max + 1);
 747         }
 748         if (entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW) {
 749                 addr_max = pernet->local_addr_max;
 750                 WRITE_ONCE(pernet->local_addr_max, addr_max + 1);
 751         }
 752
 753         pernet->addrs++;
 754         list_add_tail_rcu(&entry->list, &pernet->local_addr_list);
 755         ret = entry->addr.id;
 756
 757 out:
 758         spin_unlock_bh(&pernet->lock);
 759         return ret;
 760 }
 761
 762 static int mptcp_pm_nl_create_listen_socket(struct sock *sk,
 763                                             struct mptcp_pm_addr_entry *entry)
 764 {
 765         struct sockaddr_storage addr;
 766         struct mptcp_sock *msk;
 767         struct socket *ssock;
 768         int backlog = 1024;
 769         int err;
 770
 771         err = sock_create_kern(sock_net(sk), entry->addr.family,
 772                                SOCK_STREAM, IPPROTO_MPTCP, &entry->lsk);
 773         if (err)
 774                 return err;
 775
 776         msk = mptcp_sk(entry->lsk->sk);
 777         if (!msk) {
 778                 err = -EINVAL;
 779                 goto out;
 780         }
 781
 782         ssock = __mptcp_nmpc_socket(msk);
 783         if (!ssock) {
 784                 err = -EINVAL;
 785                 goto out;
 786         }
 787
 788         mptcp_info2sockaddr(&entry->addr, &addr, entry->addr.family);
 789         err = kernel_bind(ssock, (struct sockaddr *)&addr,
 790                           sizeof(struct sockaddr_in));
 791         if (err) {
 792                 pr_warn("kernel_bind error, err=%d", err);
 793                 goto out;
 794         }
 795
 796         err = kernel_listen(ssock, backlog);
 797         if (err) {
 798                 pr_warn("kernel_listen error, err=%d", err);
 799                 goto out;
 800         }
 801
 802         return 0;
 803
 804 out:
 805         sock_release(entry->lsk);
 806         return err;
 807 }
 808
 809 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc)
 810 {
 811         struct mptcp_pm_addr_entry *entry;
 812         struct mptcp_addr_info skc_local;
 813         struct mptcp_addr_info msk_local;
 814         struct pm_nl_pernet *pernet;
 815         int ret = -1;
 816
 817         if (WARN_ON_ONCE(!msk))
 818                 return -1;
 819
 820         /* The 0 ID mapping is defined by the first subflow, copied into the msk
 821          * addr
 822          */
 823         local_address((struct sock_common *)msk, &msk_local);
 824         local_address((struct sock_common *)skc, &skc_local);
 825         if (addresses_equal(&msk_local, &skc_local, false))
 826                 return 0;
 827
 828         if (address_zero(&skc_local))
 829                 return 0;
 830
 831         pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
 832
 833         rcu_read_lock();
 834         list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
 835                 if (addresses_equal(&entry->addr, &skc_local, entry->addr.port)) {
 836                         ret = entry->addr.id;
 837                         break;
 838                 }
 839         }
 840         rcu_read_unlock();
 841         if (ret >= 0)
 842                 return ret;
 843
 844         /* address not found, add to local list */
 845         entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
 846         if (!entry)
 847                 return -ENOMEM;
 848
 849         entry->addr = skc_local;
 850         entry->addr.id = 0;
 851         entry->addr.port = 0;
 852         entry->ifindex = 0;
 853         entry->flags = 0;
 854         entry->lsk = NULL;
 855         ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
 856         if (ret < 0)
 857                 kfree(entry);
 858
 859         return ret;
 860 }
 861
 862 void mptcp_pm_nl_data_init(struct mptcp_sock *msk)
 863 {
 864         struct mptcp_pm_data *pm = &msk->pm;
 865         bool subflows;
 866
 867         subflows = !!mptcp_pm_get_subflows_max(msk);
 868         WRITE_ONCE(pm->work_pending, (!!mptcp_pm_get_local_addr_max(msk) && subflows) ||
 869                    !!mptcp_pm_get_add_addr_signal_max(msk));
 870         WRITE_ONCE(pm->accept_addr, !!mptcp_pm_get_add_addr_accept_max(msk) && subflows);
 871         WRITE_ONCE(pm->accept_subflow, subflows);
 872 }
 873
 874 #define MPTCP_PM_CMD_GRP_OFFSET       0
 875 #define MPTCP_PM_EV_GRP_OFFSET        1
 876
 877 static const struct genl_multicast_group mptcp_pm_mcgrps[] = {
 878         [MPTCP_PM_CMD_GRP_OFFSET]       = { .name = MPTCP_PM_CMD_GRP_NAME, },
 879         [MPTCP_PM_EV_GRP_OFFSET]        = { .name = MPTCP_PM_EV_GRP_NAME,
 880                                             .flags = GENL_UNS_ADMIN_PERM,
 881                                           },
 882 };
 883
 884 static const struct nla_policy
 885 mptcp_pm_addr_policy[MPTCP_PM_ADDR_ATTR_MAX + 1] = {
 886         [MPTCP_PM_ADDR_ATTR_FAMILY]     = { .type       = NLA_U16,      },
 887         [MPTCP_PM_ADDR_ATTR_ID]         = { .type       = NLA_U8,       },
 888         [MPTCP_PM_ADDR_ATTR_ADDR4]      = { .type       = NLA_U32,      },
 889         [MPTCP_PM_ADDR_ATTR_ADDR6]      =
 890                 NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)),
 891         [MPTCP_PM_ADDR_ATTR_PORT]       = { .type       = NLA_U16       },
 892         [MPTCP_PM_ADDR_ATTR_FLAGS]      = { .type       = NLA_U32       },
 893         [MPTCP_PM_ADDR_ATTR_IF_IDX]     = { .type       = NLA_S32       },
 894 };
 895
 896 static const struct nla_policy mptcp_pm_policy[MPTCP_PM_ATTR_MAX + 1] = {
 897         [MPTCP_PM_ATTR_ADDR]            =
 898                                         NLA_POLICY_NESTED(mptcp_pm_addr_policy),
 899         [MPTCP_PM_ATTR_RCV_ADD_ADDRS]   = { .type       = NLA_U32,      },
 900         [MPTCP_PM_ATTR_SUBFLOWS]        = { .type       = NLA_U32,      },
 901 };
 902
 903 static int mptcp_pm_family_to_addr(int family)
 904 {
 905 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 906         if (family == AF_INET6)
 907                 return MPTCP_PM_ADDR_ATTR_ADDR6;
 908 #endif
 909         return MPTCP_PM_ADDR_ATTR_ADDR4;
 910 }
 911
 912 static int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
 913                                bool require_family,
 914                                struct mptcp_pm_addr_entry *entry)
 915 {
 916         struct nlattr *tb[MPTCP_PM_ADDR_ATTR_MAX + 1];
 917         int err, addr_addr;
 918
 919         if (!attr) {
 920                 GENL_SET_ERR_MSG(info, "missing address info");
 921                 return -EINVAL;
 922         }
 923
 924         /* no validation needed - was already done via nested policy */
 925         err = nla_parse_nested_deprecated(tb, MPTCP_PM_ADDR_ATTR_MAX, attr,
 926                                           mptcp_pm_addr_policy, info->extack);
 927         if (err)
 928                 return err;
 929
 930         memset(entry, 0, sizeof(*entry));
 931         if (!tb[MPTCP_PM_ADDR_ATTR_FAMILY]) {
 932                 if (!require_family)
 933                         goto skip_family;
 934
 935                 NL_SET_ERR_MSG_ATTR(info->extack, attr,
 936                                     "missing family");
 937                 return -EINVAL;
 938         }
 939
 940         entry->addr.family = nla_get_u16(tb[MPTCP_PM_ADDR_ATTR_FAMILY]);
 941         if (entry->addr.family != AF_INET
 942 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 943             && entry->addr.family != AF_INET6
 944 #endif
 945             ) {
 946                 NL_SET_ERR_MSG_ATTR(info->extack, attr,
 947                                     "unknown address family");
 948                 return -EINVAL;
 949         }
 950         addr_addr = mptcp_pm_family_to_addr(entry->addr.family);
 951         if (!tb[addr_addr]) {
 952                 NL_SET_ERR_MSG_ATTR(info->extack, attr,
 953                                     "missing address data");
 954                 return -EINVAL;
 955         }
 956
 957 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 958         if (entry->addr.family == AF_INET6)
 959                 entry->addr.addr6 = nla_get_in6_addr(tb[addr_addr]);
 960         else
 961 #endif
 962                 entry->addr.addr.s_addr = nla_get_in_addr(tb[addr_addr]);
 963
 964 skip_family:
 965         if (tb[MPTCP_PM_ADDR_ATTR_IF_IDX]) {
 966                 u32 val = nla_get_s32(tb[MPTCP_PM_ADDR_ATTR_IF_IDX]);
 967
 968                 entry->ifindex = val;
 969         }
 970
 971         if (tb[MPTCP_PM_ADDR_ATTR_ID])
 972                 entry->addr.id = nla_get_u8(tb[MPTCP_PM_ADDR_ATTR_ID]);
 973
 974         if (tb[MPTCP_PM_ADDR_ATTR_FLAGS])
 975                 entry->flags = nla_get_u32(tb[MPTCP_PM_ADDR_ATTR_FLAGS]);
 976
 977         if (tb[MPTCP_PM_ADDR_ATTR_PORT]) {
 978                 if (!(entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL)) {
 979                         NL_SET_ERR_MSG_ATTR(info->extack, attr,
 980                                             "flags must have signal when using port");
 981                         return -EINVAL;
 982                 }
 983                 entry->addr.port = htons(nla_get_u16(tb[MPTCP_PM_ADDR_ATTR_PORT]));
 984         }
 985
 986         return 0;
 987 }
 988
 989 static struct pm_nl_pernet *genl_info_pm_nl(struct genl_info *info)
 990 {
 991         return net_generic(genl_info_net(info), pm_nl_pernet_id);
 992 }
 993
 994 static int mptcp_nl_add_subflow_or_signal_addr(struct net *net)
 995 {
 996         struct mptcp_sock *msk;
 997         long s_slot = 0, s_num = 0;
 998
 999         while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1000                 struct sock *sk = (struct sock *)msk;
1001
1002                 if (!READ_ONCE(msk->fully_established))
1003                         goto next;
1004
1005                 lock_sock(sk);
1006                 spin_lock_bh(&msk->pm.lock);
1007                 mptcp_pm_create_subflow_or_signal_addr(msk);
1008                 spin_unlock_bh(&msk->pm.lock);
1009                 release_sock(sk);
1010
1011 next:
1012                 sock_put(sk);
1013                 cond_resched();
1014         }
1015
1016         return 0;
1017 }
1018
1019 static int mptcp_nl_cmd_add_addr(struct sk_buff *skb, struct genl_info *info)
1020 {
1021         struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1022         struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1023         struct mptcp_pm_addr_entry addr, *entry;
1024         int ret;
1025
1026         ret = mptcp_pm_parse_addr(attr, info, true, &addr);
1027         if (ret < 0)
1028                 return ret;
1029
1030         entry = kmalloc(sizeof(*entry), GFP_KERNEL);
1031         if (!entry) {
1032                 GENL_SET_ERR_MSG(info, "can't allocate addr");
1033                 return -ENOMEM;
1034         }
1035
1036         *entry = addr;
1037         if (entry->addr.port) {
1038                 ret = mptcp_pm_nl_create_listen_socket(skb->sk, entry);
1039                 if (ret) {
1040                         GENL_SET_ERR_MSG(info, "create listen socket error");
1041                         kfree(entry);
1042                         return ret;
1043                 }
1044         }
1045         ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
1046         if (ret < 0) {
1047                 GENL_SET_ERR_MSG(info, "too many addresses or duplicate one");
1048                 if (entry->lsk)
1049                         sock_release(entry->lsk);
1050                 kfree(entry);
1051                 return ret;
1052         }
1053
1054         mptcp_nl_add_subflow_or_signal_addr(sock_net(skb->sk));
1055
1056         return 0;
1057 }
1058
1059 static struct mptcp_pm_addr_entry *
1060 __lookup_addr_by_id(struct pm_nl_pernet *pernet, unsigned int id)
1061 {
1062         struct mptcp_pm_addr_entry *entry;
1063
1064         list_for_each_entry(entry, &pernet->local_addr_list, list) {
1065                 if (entry->addr.id == id)
1066                         return entry;
1067         }
1068         return NULL;
1069 }
1070
1071 static bool remove_anno_list_by_saddr(struct mptcp_sock *msk,
1072                                       struct mptcp_addr_info *addr)
1073 {
1074         struct mptcp_pm_add_entry *entry;
1075
1076         entry = mptcp_pm_del_add_timer(msk, addr, false);
1077         if (entry) {
1078                 list_del(&entry->list);
1079                 kfree(entry);
1080                 return true;
1081         }
1082
1083         return false;
1084 }
1085
1086 static bool mptcp_pm_remove_anno_addr(struct mptcp_sock *msk,
1087                                       struct mptcp_addr_info *addr,
1088                                       bool force)
1089 {
1090         struct mptcp_rm_list list = { .nr = 0 };
1091         bool ret;
1092
1093         list.ids[list.nr++] = addr->id;
1094
1095         ret = remove_anno_list_by_saddr(msk, addr);
1096         if (ret || force) {
1097                 spin_lock_bh(&msk->pm.lock);
1098                 mptcp_pm_remove_addr(msk, &list);
1099                 spin_unlock_bh(&msk->pm.lock);
1100         }
1101         return ret;
1102 }
1103
1104 static int mptcp_nl_remove_subflow_and_signal_addr(struct net *net,
1105                                                    struct mptcp_addr_info *addr)
1106 {
1107         struct mptcp_sock *msk;
1108         long s_slot = 0, s_num = 0;
1109         struct mptcp_rm_list list = { .nr = 0 };
1110
1111         pr_debug("remove_id=%d", addr->id);
1112
1113         list.ids[list.nr++] = addr->id;
1114
1115         while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1116                 struct sock *sk = (struct sock *)msk;
1117                 bool remove_subflow;
1118
1119                 if (list_empty(&msk->conn_list)) {
1120                         mptcp_pm_remove_anno_addr(msk, addr, false);
1121                         goto next;
1122                 }
1123
1124                 lock_sock(sk);
1125                 remove_subflow = lookup_subflow_by_saddr(&msk->conn_list, addr);
1126                 mptcp_pm_remove_anno_addr(msk, addr, remove_subflow);
1127                 if (remove_subflow)
1128                         mptcp_pm_remove_subflow(msk, &list);
1129                 release_sock(sk);
1130
1131 next:
1132                 sock_put(sk);
1133                 cond_resched();
1134         }
1135
1136         return 0;
1137 }
1138
1139 struct addr_entry_release_work {
1140         struct rcu_work rwork;
1141         struct mptcp_pm_addr_entry *entry;
1142 };
1143
1144 static void mptcp_pm_release_addr_entry(struct work_struct *work)
1145 {
1146         struct addr_entry_release_work *w;
1147         struct mptcp_pm_addr_entry *entry;
1148
1149         w = container_of(to_rcu_work(work), struct addr_entry_release_work, rwork);
1150         entry = w->entry;
1151         if (entry) {
1152                 if (entry->lsk)
1153                         sock_release(entry->lsk);
1154                 kfree(entry);
1155         }
1156         kfree(w);
1157 }
1158
1159 static void mptcp_pm_free_addr_entry(struct mptcp_pm_addr_entry *entry)
1160 {
1161         struct addr_entry_release_work *w;
1162
1163         w = kmalloc(sizeof(*w), GFP_ATOMIC);
1164         if (w) {
1165                 INIT_RCU_WORK(&w->rwork, mptcp_pm_release_addr_entry);
1166                 w->entry = entry;
1167                 queue_rcu_work(system_wq, &w->rwork);
1168         }
1169 }
1170
1171 static int mptcp_nl_remove_id_zero_address(struct net *net,
1172                                            struct mptcp_addr_info *addr)
1173 {
1174         struct mptcp_rm_list list = { .nr = 0 };
1175         long s_slot = 0, s_num = 0;
1176         struct mptcp_sock *msk;
1177
1178         list.ids[list.nr++] = 0;
1179
1180         while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1181                 struct sock *sk = (struct sock *)msk;
1182                 struct mptcp_addr_info msk_local;
1183
1184                 if (list_empty(&msk->conn_list))
1185                         goto next;
1186
1187                 local_address((struct sock_common *)msk, &msk_local);
1188                 if (!addresses_equal(&msk_local, addr, addr->port))
1189                         goto next;
1190
1191                 lock_sock(sk);
1192                 spin_lock_bh(&msk->pm.lock);
1193                 mptcp_pm_remove_addr(msk, &list);
1194                 mptcp_pm_nl_rm_subflow_received(msk, &list);
1195                 spin_unlock_bh(&msk->pm.lock);
1196                 release_sock(sk);
1197
1198 next:
1199                 sock_put(sk);
1200                 cond_resched();
1201         }
1202
1203         return 0;
1204 }
1205
1206 static int mptcp_nl_cmd_del_addr(struct sk_buff *skb, struct genl_info *info)
1207 {
1208         struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1209         struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1210         struct mptcp_pm_addr_entry addr, *entry;
1211         unsigned int addr_max;
1212         int ret;
1213
1214         ret = mptcp_pm_parse_addr(attr, info, false, &addr);
1215         if (ret < 0)
1216                 return ret;
1217
1218         /* the zero id address is special: the first address used by the msk
1219          * always gets such an id, so different subflows can have different zero
1220          * id addresses. Additionally zero id is not accounted for in id_bitmap.
1221          * Let's use an 'mptcp_rm_list' instead of the common remove code.
1222          */
1223         if (addr.addr.id == 0)
1224                 return mptcp_nl_remove_id_zero_address(sock_net(skb->sk), &addr.addr);
1225
1226         spin_lock_bh(&pernet->lock);
1227         entry = __lookup_addr_by_id(pernet, addr.addr.id);
1228         if (!entry) {
1229                 GENL_SET_ERR_MSG(info, "address not found");
1230                 spin_unlock_bh(&pernet->lock);
1231                 return -EINVAL;
1232         }
1233         if (entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL) {
1234                 addr_max = pernet->add_addr_signal_max;
1235                 WRITE_ONCE(pernet->add_addr_signal_max, addr_max - 1);
1236         }
1237         if (entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW) {
1238                 addr_max = pernet->local_addr_max;
1239                 WRITE_ONCE(pernet->local_addr_max, addr_max - 1);
1240         }
1241
1242         pernet->addrs--;
1243         list_del_rcu(&entry->list);
1244         __clear_bit(entry->addr.id, pernet->id_bitmap);
1245         spin_unlock_bh(&pernet->lock);
1246
1247         mptcp_nl_remove_subflow_and_signal_addr(sock_net(skb->sk), &entry->addr);
1248         mptcp_pm_free_addr_entry(entry);
1249
1250         return ret;
1251 }
1252
1253 static void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk,
1254                                                struct list_head *rm_list)
1255 {
1256         struct mptcp_rm_list alist = { .nr = 0 }, slist = { .nr = 0 };
1257         struct mptcp_pm_addr_entry *entry;
1258
1259         list_for_each_entry(entry, rm_list, list) {
1260                 if (lookup_subflow_by_saddr(&msk->conn_list, &entry->addr) &&
1261                     alist.nr < MPTCP_RM_IDS_MAX &&
1262                     slist.nr < MPTCP_RM_IDS_MAX) {
1263                         alist.ids[alist.nr++] = entry->addr.id;
1264                         slist.ids[slist.nr++] = entry->addr.id;
1265                 } else if (remove_anno_list_by_saddr(msk, &entry->addr) &&
1266                          alist.nr < MPTCP_RM_IDS_MAX) {
1267                         alist.ids[alist.nr++] = entry->addr.id;
1268                 }
1269         }
1270
1271         if (alist.nr) {
1272                 spin_lock_bh(&msk->pm.lock);
1273                 mptcp_pm_remove_addr(msk, &alist);
1274                 spin_unlock_bh(&msk->pm.lock);
1275         }
1276         if (slist.nr)
1277                 mptcp_pm_remove_subflow(msk, &slist);
1278 }
1279
1280 static void mptcp_nl_remove_addrs_list(struct net *net,
1281                                        struct list_head *rm_list)
1282 {
1283         long s_slot = 0, s_num = 0;
1284         struct mptcp_sock *msk;
1285
1286         if (list_empty(rm_list))
1287                 return;
1288
1289         while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1290                 struct sock *sk = (struct sock *)msk;
1291
1292                 lock_sock(sk);
1293                 mptcp_pm_remove_addrs_and_subflows(msk, rm_list);
1294                 release_sock(sk);
1295
1296                 sock_put(sk);
1297                 cond_resched();
1298         }
1299 }
1300
1301 static void __flush_addrs(struct list_head *list)
1302 {
1303         while (!list_empty(list)) {
1304                 struct mptcp_pm_addr_entry *cur;
1305
1306                 cur = list_entry(list->next,
1307                                  struct mptcp_pm_addr_entry, list);
1308                 list_del_rcu(&cur->list);
1309                 mptcp_pm_free_addr_entry(cur);
1310         }
1311 }
1312
1313 static void __reset_counters(struct pm_nl_pernet *pernet)
1314 {
1315         WRITE_ONCE(pernet->add_addr_signal_max, 0);
1316         WRITE_ONCE(pernet->add_addr_accept_max, 0);
1317         WRITE_ONCE(pernet->local_addr_max, 0);
1318         pernet->addrs = 0;
1319 }
1320
1321 static int mptcp_nl_cmd_flush_addrs(struct sk_buff *skb, struct genl_info *info)
1322 {
1323         struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1324         LIST_HEAD(free_list);
1325
1326         spin_lock_bh(&pernet->lock);
1327         list_splice_init(&pernet->local_addr_list, &free_list);
1328         __reset_counters(pernet);
1329         pernet->next_id = 1;
1330         bitmap_zero(pernet->id_bitmap, MAX_ADDR_ID + 1);
1331         spin_unlock_bh(&pernet->lock);
1332         mptcp_nl_remove_addrs_list(sock_net(skb->sk), &free_list);
1333         __flush_addrs(&free_list);
1334         return 0;
1335 }
1336
1337 static int mptcp_nl_fill_addr(struct sk_buff *skb,
1338                               struct mptcp_pm_addr_entry *entry)
1339 {
1340         struct mptcp_addr_info *addr = &entry->addr;
1341         struct nlattr *attr;
1342
1343         attr = nla_nest_start(skb, MPTCP_PM_ATTR_ADDR);
1344         if (!attr)
1345                 return -EMSGSIZE;
1346
1347         if (nla_put_u16(skb, MPTCP_PM_ADDR_ATTR_FAMILY, addr->family))
1348                 goto nla_put_failure;
1349         if (nla_put_u16(skb, MPTCP_PM_ADDR_ATTR_PORT, ntohs(addr->port)))
1350                 goto nla_put_failure;
1351         if (nla_put_u8(skb, MPTCP_PM_ADDR_ATTR_ID, addr->id))
1352                 goto nla_put_failure;
1353         if (nla_put_u32(skb, MPTCP_PM_ADDR_ATTR_FLAGS, entry->flags))
1354                 goto nla_put_failure;
1355         if (entry->ifindex &&
1356             nla_put_s32(skb, MPTCP_PM_ADDR_ATTR_IF_IDX, entry->ifindex))
1357                 goto nla_put_failure;
1358
1359         if (addr->family == AF_INET &&
1360             nla_put_in_addr(skb, MPTCP_PM_ADDR_ATTR_ADDR4,
1361                             addr->addr.s_addr))
1362                 goto nla_put_failure;
1363 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1364         else if (addr->family == AF_INET6 &&
1365                  nla_put_in6_addr(skb, MPTCP_PM_ADDR_ATTR_ADDR6, &addr->addr6))
1366                 goto nla_put_failure;
1367 #endif
1368         nla_nest_end(skb, attr);
1369         return 0;
1370
1371 nla_put_failure:
1372         nla_nest_cancel(skb, attr);
1373         return -EMSGSIZE;
1374 }
1375
1376 static int mptcp_nl_cmd_get_addr(struct sk_buff *skb, struct genl_info *info)
1377 {
1378         struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1379         struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1380         struct mptcp_pm_addr_entry addr, *entry;
1381         struct sk_buff *msg;
1382         void *reply;
1383         int ret;
1384
1385         ret = mptcp_pm_parse_addr(attr, info, false, &addr);
1386         if (ret < 0)
1387                 return ret;
1388
1389         msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1390         if (!msg)
1391                 return -ENOMEM;
1392
1393         reply = genlmsg_put_reply(msg, info, &mptcp_genl_family, 0,
1394                                   info->genlhdr->cmd);
1395         if (!reply) {
1396                 GENL_SET_ERR_MSG(info, "not enough space in Netlink message");
1397                 ret = -EMSGSIZE;
1398                 goto fail;
1399         }
1400
1401         spin_lock_bh(&pernet->lock);
1402         entry = __lookup_addr_by_id(pernet, addr.addr.id);
1403         if (!entry) {
1404                 GENL_SET_ERR_MSG(info, "address not found");
1405                 ret = -EINVAL;
1406                 goto unlock_fail;
1407         }
1408
1409         ret = mptcp_nl_fill_addr(msg, entry);
1410         if (ret)
1411                 goto unlock_fail;
1412
1413         genlmsg_end(msg, reply);
1414         ret = genlmsg_reply(msg, info);
1415         spin_unlock_bh(&pernet->lock);
1416         return ret;
1417
1418 unlock_fail:
1419         spin_unlock_bh(&pernet->lock);
1420
1421 fail:
1422         nlmsg_free(msg);
1423         return ret;
1424 }
1425
1426 static int mptcp_nl_cmd_dump_addrs(struct sk_buff *msg,
1427                                    struct netlink_callback *cb)
1428 {
1429         struct net *net = sock_net(msg->sk);
1430         struct mptcp_pm_addr_entry *entry;
1431         struct pm_nl_pernet *pernet;
1432         int id = cb->args[0];
1433         void *hdr;
1434         int i;
1435
1436         pernet = net_generic(net, pm_nl_pernet_id);
1437
1438         spin_lock_bh(&pernet->lock);
1439         for (i = id; i < MAX_ADDR_ID + 1; i++) {
1440                 if (test_bit(i, pernet->id_bitmap)) {
1441                         entry = __lookup_addr_by_id(pernet, i);
1442                         if (!entry)
1443                                 break;
1444
1445                         if (entry->addr.id <= id)
1446                                 continue;
1447
1448                         hdr = genlmsg_put(msg, NETLINK_CB(cb->skb).portid,
1449                                           cb->nlh->nlmsg_seq, &mptcp_genl_family,
1450                                           NLM_F_MULTI, MPTCP_PM_CMD_GET_ADDR);
1451                         if (!hdr)
1452                                 break;
1453
1454                         if (mptcp_nl_fill_addr(msg, entry) < 0) {
1455                                 genlmsg_cancel(msg, hdr);
1456                                 break;
1457                         }
1458
1459                         id = entry->addr.id;
1460                         genlmsg_end(msg, hdr);
1461                 }
1462         }
1463         spin_unlock_bh(&pernet->lock);
1464
1465         cb->args[0] = id;
1466         return msg->len;
1467 }
1468
1469 static int parse_limit(struct genl_info *info, int id, unsigned int *limit)
1470 {
1471         struct nlattr *attr = info->attrs[id];
1472
1473         if (!attr)
1474                 return 0;
1475
1476         *limit = nla_get_u32(attr);
1477         if (*limit > MPTCP_PM_ADDR_MAX) {
1478                 GENL_SET_ERR_MSG(info, "limit greater than maximum");
1479                 return -EINVAL;
1480         }
1481         return 0;
1482 }
1483
1484 static int
1485 mptcp_nl_cmd_set_limits(struct sk_buff *skb, struct genl_info *info)
1486 {
1487         struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1488         unsigned int rcv_addrs, subflows;
1489         int ret;
1490
1491         spin_lock_bh(&pernet->lock);
1492         rcv_addrs = pernet->add_addr_accept_max;
1493         ret = parse_limit(info, MPTCP_PM_ATTR_RCV_ADD_ADDRS, &rcv_addrs);
1494         if (ret)
1495                 goto unlock;
1496
1497         subflows = pernet->subflows_max;
1498         ret = parse_limit(info, MPTCP_PM_ATTR_SUBFLOWS, &subflows);
1499         if (ret)
1500                 goto unlock;
1501
1502         WRITE_ONCE(pernet->add_addr_accept_max, rcv_addrs);
1503         WRITE_ONCE(pernet->subflows_max, subflows);
1504
1505 unlock:
1506         spin_unlock_bh(&pernet->lock);
1507         return ret;
1508 }
1509
1510 static int
1511 mptcp_nl_cmd_get_limits(struct sk_buff *skb, struct genl_info *info)
1512 {
1513         struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1514         struct sk_buff *msg;
1515         void *reply;
1516
1517         msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1518         if (!msg)
1519                 return -ENOMEM;
1520
1521         reply = genlmsg_put_reply(msg, info, &mptcp_genl_family, 0,
1522                                   MPTCP_PM_CMD_GET_LIMITS);
1523         if (!reply)
1524                 goto fail;
1525
1526         if (nla_put_u32(msg, MPTCP_PM_ATTR_RCV_ADD_ADDRS,
1527                         READ_ONCE(pernet->add_addr_accept_max)))
1528                 goto fail;
1529
1530         if (nla_put_u32(msg, MPTCP_PM_ATTR_SUBFLOWS,
1531                         READ_ONCE(pernet->subflows_max)))
1532                 goto fail;
1533
1534         genlmsg_end(msg, reply);
1535         return genlmsg_reply(msg, info);
1536
1537 fail:
1538         GENL_SET_ERR_MSG(info, "not enough space in Netlink message");
1539         nlmsg_free(msg);
1540         return -EMSGSIZE;
1541 }
1542
1543 static int mptcp_nl_addr_backup(struct net *net,
1544                                 struct mptcp_addr_info *addr,
1545                                 u8 bkup)
1546 {
1547         long s_slot = 0, s_num = 0;
1548         struct mptcp_sock *msk;
1549         int ret = -EINVAL;
1550
1551         while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1552                 struct sock *sk = (struct sock *)msk;
1553
1554                 if (list_empty(&msk->conn_list))
1555                         goto next;
1556
1557                 lock_sock(sk);
1558                 spin_lock_bh(&msk->pm.lock);
1559                 ret = mptcp_pm_nl_mp_prio_send_ack(msk, addr, bkup);
1560                 spin_unlock_bh(&msk->pm.lock);
1561                 release_sock(sk);
1562
1563 next:
1564                 sock_put(sk);
1565                 cond_resched();
1566         }
1567
1568         return ret;
1569 }
1570
1571 static int mptcp_nl_cmd_set_flags(struct sk_buff *skb, struct genl_info *info)
1572 {
1573         struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1574         struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1575         struct mptcp_pm_addr_entry addr, *entry;
1576         struct net *net = sock_net(skb->sk);
1577         u8 bkup = 0;
1578         int ret;
1579
1580         ret = mptcp_pm_parse_addr(attr, info, true, &addr);
1581         if (ret < 0)
1582                 return ret;
1583
1584         if (addr.flags & MPTCP_PM_ADDR_FLAG_BACKUP)
1585                 bkup = 1;
1586
1587         list_for_each_entry(entry, &pernet->local_addr_list, list) {
1588                 if (addresses_equal(&entry->addr, &addr.addr, true)) {
1589                         ret = mptcp_nl_addr_backup(net, &entry->addr, bkup);
1590                         if (ret)
1591                                 return ret;
1592
1593                         if (bkup)
1594                                 entry->flags |= MPTCP_PM_ADDR_FLAG_BACKUP;
1595                         else
1596                                 entry->flags &= ~MPTCP_PM_ADDR_FLAG_BACKUP;
1597                 }
1598         }
1599
1600         return 0;
1601 }
1602
1603 static void mptcp_nl_mcast_send(struct net *net, struct sk_buff *nlskb, gfp_t gfp)
1604 {
1605         genlmsg_multicast_netns(&mptcp_genl_family, net,
1606                                 nlskb, 0, MPTCP_PM_EV_GRP_OFFSET, gfp);
1607 }
1608
1609 static int mptcp_event_add_subflow(struct sk_buff *skb, const struct sock *ssk)
1610 {
1611         const struct inet_sock *issk = inet_sk(ssk);
1612         const struct mptcp_subflow_context *sf;
1613
1614         if (nla_put_u16(skb, MPTCP_ATTR_FAMILY, ssk->sk_family))
1615                 return -EMSGSIZE;
1616
1617         switch (ssk->sk_family) {
1618         case AF_INET:
1619                 if (nla_put_in_addr(skb, MPTCP_ATTR_SADDR4, issk->inet_saddr))
1620                         return -EMSGSIZE;
1621                 if (nla_put_in_addr(skb, MPTCP_ATTR_DADDR4, issk->inet_daddr))
1622                         return -EMSGSIZE;
1623                 break;
1624 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1625         case AF_INET6: {
1626                 const struct ipv6_pinfo *np = inet6_sk(ssk);
1627
1628                 if (nla_put_in6_addr(skb, MPTCP_ATTR_SADDR6, &np->saddr))
1629                         return -EMSGSIZE;
1630                 if (nla_put_in6_addr(skb, MPTCP_ATTR_DADDR6, &ssk->sk_v6_daddr))
1631                         return -EMSGSIZE;
1632                 break;
1633         }
1634 #endif
1635         default:
1636                 WARN_ON_ONCE(1);
1637                 return -EMSGSIZE;
1638         }
1639
1640         if (nla_put_be16(skb, MPTCP_ATTR_SPORT, issk->inet_sport))
1641                 return -EMSGSIZE;
1642         if (nla_put_be16(skb, MPTCP_ATTR_DPORT, issk->inet_dport))
1643                 return -EMSGSIZE;
1644
1645         sf = mptcp_subflow_ctx(ssk);
1646         if (WARN_ON_ONCE(!sf))
1647                 return -EINVAL;
1648
1649         if (nla_put_u8(skb, MPTCP_ATTR_LOC_ID, sf->local_id))
1650                 return -EMSGSIZE;
1651
1652         if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, sf->remote_id))
1653                 return -EMSGSIZE;
1654
1655         return 0;
1656 }
1657
1658 static int mptcp_event_put_token_and_ssk(struct sk_buff *skb,
1659                                          const struct mptcp_sock *msk,
1660                                          const struct sock *ssk)
1661 {
1662         const struct sock *sk = (const struct sock *)msk;
1663         const struct mptcp_subflow_context *sf;
1664         u8 sk_err;
1665
1666         if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1667                 return -EMSGSIZE;
1668
1669         if (mptcp_event_add_subflow(skb, ssk))
1670                 return -EMSGSIZE;
1671
1672         sf = mptcp_subflow_ctx(ssk);
1673         if (WARN_ON_ONCE(!sf))
1674                 return -EINVAL;
1675
1676         if (nla_put_u8(skb, MPTCP_ATTR_BACKUP, sf->backup))
1677                 return -EMSGSIZE;
1678
1679         if (ssk->sk_bound_dev_if &&
1680             nla_put_s32(skb, MPTCP_ATTR_IF_IDX, ssk->sk_bound_dev_if))
1681                 return -EMSGSIZE;
1682
1683         sk_err = ssk->sk_err;
1684         if (sk_err && sk->sk_state == TCP_ESTABLISHED &&
1685             nla_put_u8(skb, MPTCP_ATTR_ERROR, sk_err))
1686                 return -EMSGSIZE;
1687
1688         return 0;
1689 }
1690
1691 static int mptcp_event_sub_established(struct sk_buff *skb,
1692                                        const struct mptcp_sock *msk,
1693                                        const struct sock *ssk)
1694 {
1695         return mptcp_event_put_token_and_ssk(skb, msk, ssk);
1696 }
1697
1698 static int mptcp_event_sub_closed(struct sk_buff *skb,
1699                                   const struct mptcp_sock *msk,
1700                                   const struct sock *ssk)
1701 {
1702         const struct mptcp_subflow_context *sf;
1703
1704         if (mptcp_event_put_token_and_ssk(skb, msk, ssk))
1705                 return -EMSGSIZE;
1706
1707         sf = mptcp_subflow_ctx(ssk);
1708         if (!sf->reset_seen)
1709                 return 0;
1710
1711         if (nla_put_u32(skb, MPTCP_ATTR_RESET_REASON, sf->reset_reason))
1712                 return -EMSGSIZE;
1713
1714         if (nla_put_u32(skb, MPTCP_ATTR_RESET_FLAGS, sf->reset_transient))
1715                 return -EMSGSIZE;
1716
1717         return 0;
1718 }
1719
1720 static int mptcp_event_created(struct sk_buff *skb,
1721                                const struct mptcp_sock *msk,
1722                                const struct sock *ssk)
1723 {
1724         int err = nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token);
1725
1726         if (err)
1727                 return err;
1728
1729         return mptcp_event_add_subflow(skb, ssk);
1730 }
1731
1732 void mptcp_event_addr_removed(const struct mptcp_sock *msk, uint8_t id)
1733 {
1734         struct net *net = sock_net((const struct sock *)msk);
1735         struct nlmsghdr *nlh;
1736         struct sk_buff *skb;
1737
1738         if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1739                 return;
1740
1741         skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
1742         if (!skb)
1743                 return;
1744
1745         nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0, MPTCP_EVENT_REMOVED);
1746         if (!nlh)
1747                 goto nla_put_failure;
1748
1749         if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1750                 goto nla_put_failure;
1751
1752         if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, id))
1753                 goto nla_put_failure;
1754
1755         genlmsg_end(skb, nlh);
1756         mptcp_nl_mcast_send(net, skb, GFP_ATOMIC);
1757         return;
1758
1759 nla_put_failure:
1760         kfree_skb(skb);
1761 }
1762
1763 void mptcp_event_addr_announced(const struct mptcp_sock *msk,
1764                                 const struct mptcp_addr_info *info)
1765 {
1766         struct net *net = sock_net((const struct sock *)msk);
1767         struct nlmsghdr *nlh;
1768         struct sk_buff *skb;
1769
1770         if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1771                 return;
1772
1773         skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
1774         if (!skb)
1775                 return;
1776
1777         nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0,
1778                           MPTCP_EVENT_ANNOUNCED);
1779         if (!nlh)
1780                 goto nla_put_failure;
1781
1782         if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1783                 goto nla_put_failure;
1784
1785         if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, info->id))
1786                 goto nla_put_failure;
1787
1788         if (nla_put_be16(skb, MPTCP_ATTR_DPORT, info->port))
1789                 goto nla_put_failure;
1790
1791         switch (info->family) {
1792         case AF_INET:
1793                 if (nla_put_in_addr(skb, MPTCP_ATTR_DADDR4, info->addr.s_addr))
1794                         goto nla_put_failure;
1795                 break;
1796 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1797         case AF_INET6:
1798                 if (nla_put_in6_addr(skb, MPTCP_ATTR_DADDR6, &info->addr6))
1799                         goto nla_put_failure;
1800                 break;
1801 #endif
1802         default:
1803                 WARN_ON_ONCE(1);
1804                 goto nla_put_failure;
1805         }
1806
1807         genlmsg_end(skb, nlh);
1808         mptcp_nl_mcast_send(net, skb, GFP_ATOMIC);
1809         return;
1810
1811 nla_put_failure:
1812         kfree_skb(skb);
1813 }
1814
1815 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
1816                  const struct sock *ssk, gfp_t gfp)
1817 {
1818         struct net *net = sock_net((const struct sock *)msk);
1819         struct nlmsghdr *nlh;
1820         struct sk_buff *skb;
1821
1822         if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1823                 return;
1824
1825         skb = nlmsg_new(NLMSG_DEFAULT_SIZE, gfp);
1826         if (!skb)
1827                 return;
1828
1829         nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0, type);
1830         if (!nlh)
1831                 goto nla_put_failure;
1832
1833         switch (type) {
1834         case MPTCP_EVENT_UNSPEC:
1835                 WARN_ON_ONCE(1);
1836                 break;
1837         case MPTCP_EVENT_CREATED:
1838         case MPTCP_EVENT_ESTABLISHED:
1839                 if (mptcp_event_created(skb, msk, ssk) < 0)
1840                         goto nla_put_failure;
1841                 break;
1842         case MPTCP_EVENT_CLOSED:
1843                 if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token) < 0)
1844                         goto nla_put_failure;
1845                 break;
1846         case MPTCP_EVENT_ANNOUNCED:
1847         case MPTCP_EVENT_REMOVED:
1848                 /* call mptcp_event_addr_announced()/removed instead */
1849                 WARN_ON_ONCE(1);
1850                 break;
1851         case MPTCP_EVENT_SUB_ESTABLISHED:
1852         case MPTCP_EVENT_SUB_PRIORITY:
1853                 if (mptcp_event_sub_established(skb, msk, ssk) < 0)
1854                         goto nla_put_failure;
1855                 break;
1856         case MPTCP_EVENT_SUB_CLOSED:
1857                 if (mptcp_event_sub_closed(skb, msk, ssk) < 0)
1858                         goto nla_put_failure;
1859                 break;
1860         }
1861
1862         genlmsg_end(skb, nlh);
1863         mptcp_nl_mcast_send(net, skb, gfp);
1864         return;
1865
1866 nla_put_failure:
1867         kfree_skb(skb);
1868 }
1869
1870 static const struct genl_small_ops mptcp_pm_ops[] = {
1871         {
1872                 .cmd    = MPTCP_PM_CMD_ADD_ADDR,
1873                 .doit   = mptcp_nl_cmd_add_addr,
1874                 .flags  = GENL_ADMIN_PERM,
1875         },
1876         {
1877                 .cmd    = MPTCP_PM_CMD_DEL_ADDR,
1878                 .doit   = mptcp_nl_cmd_del_addr,
1879                 .flags  = GENL_ADMIN_PERM,
1880         },
1881         {
1882                 .cmd    = MPTCP_PM_CMD_FLUSH_ADDRS,
1883                 .doit   = mptcp_nl_cmd_flush_addrs,
1884                 .flags  = GENL_ADMIN_PERM,
1885         },
1886         {
1887                 .cmd    = MPTCP_PM_CMD_GET_ADDR,
1888                 .doit   = mptcp_nl_cmd_get_addr,
1889                 .dumpit   = mptcp_nl_cmd_dump_addrs,
1890         },
1891         {
1892                 .cmd    = MPTCP_PM_CMD_SET_LIMITS,
1893                 .doit   = mptcp_nl_cmd_set_limits,
1894                 .flags  = GENL_ADMIN_PERM,
1895         },
1896         {
1897                 .cmd    = MPTCP_PM_CMD_GET_LIMITS,
1898                 .doit   = mptcp_nl_cmd_get_limits,
1899         },
1900         {
1901                 .cmd    = MPTCP_PM_CMD_SET_FLAGS,
1902                 .doit   = mptcp_nl_cmd_set_flags,
1903                 .flags  = GENL_ADMIN_PERM,
1904         },
1905 };
1906
1907 static struct genl_family mptcp_genl_family __ro_after_init = {
1908         .name           = MPTCP_PM_NAME,
1909         .version        = MPTCP_PM_VER,
1910         .maxattr        = MPTCP_PM_ATTR_MAX,
1911         .policy         = mptcp_pm_policy,
1912         .netnsok        = true,
1913         .module         = THIS_MODULE,
1914         .small_ops      = mptcp_pm_ops,
1915         .n_small_ops    = ARRAY_SIZE(mptcp_pm_ops),
1916         .mcgrps         = mptcp_pm_mcgrps,
1917         .n_mcgrps       = ARRAY_SIZE(mptcp_pm_mcgrps),
1918 };
1919
1920 static int __net_init pm_nl_init_net(struct net *net)
1921 {
1922         struct pm_nl_pernet *pernet = net_generic(net, pm_nl_pernet_id);
1923
1924         INIT_LIST_HEAD_RCU(&pernet->local_addr_list);
1925         pernet->next_id = 1;
1926         spin_lock_init(&pernet->lock);
1927
1928         /* No need to initialize other pernet fields, the struct is zeroed at
1929          * allocation time.
1930          */
1931
1932         return 0;
1933 }
1934
1935 static void __net_exit pm_nl_exit_net(struct list_head *net_list)
1936 {
1937         struct net *net;
1938
1939         list_for_each_entry(net, net_list, exit_list) {
1940                 struct pm_nl_pernet *pernet = net_generic(net, pm_nl_pernet_id);
1941
1942                 /* net is removed from namespace list, can't race with
1943                  * other modifiers
1944                  */
1945                 __flush_addrs(&pernet->local_addr_list);
1946         }
1947 }
1948
1949 static struct pernet_operations mptcp_pm_pernet_ops = {
1950         .init = pm_nl_init_net,
1951         .exit_batch = pm_nl_exit_net,
1952         .id = &pm_nl_pernet_id,
1953         .size = sizeof(struct pm_nl_pernet),
1954 };
1955
1956 void __init mptcp_pm_nl_init(void)
1957 {
1958         if (register_pernet_subsys(&mptcp_pm_pernet_ops) < 0)
1959                 panic("Failed to register MPTCP PM pernet subsystem.\n");
1960
1961         if (genl_register_family(&mptcp_genl_family))
1962                 panic("Failed to register MPTCP PM netlink family\n");
1963 }